Ibis validation > Forums > www.cdnusers.org

Forums

Subject: Ibis validation

Posting to forums is available to community members only.
Login or Register

Rate this topic:

Author

Messages

darmoni
Posts: 8
Online: User is Offline

11/15/2007 5:12 AM
Hello, i try to understand the method for validating the ibis model that i get from the vendor. i tried to correlate the simulations results vs. the scope wave from the actual board but the correlation is poor (using the SQuest & SigExp). what are the steps for ibis validating & correlation? thank you, Doron.

Kalevi2
Moderator
Posts: 69
Online: User is Offline

11/15/2007 5:49 AM
Your question requires a long answer. First, is your measurement set up using the appropriate bandwidth to be able to properly see the waveform? We have had cases where overshoots and undershoots are missed due to insufficient bandwidth. Do you have a low inductance ground path for the scope probe? Are you using a high impedance probe or making a coaxial measurement? Is the board file correctly set up for simulation? Have you checked that the IBIS file has a reasonable package model? Does the ibis file pass the ibis parser and what kind of warnings do you get? Can the ibis buffer drive into the reference load at the max specified frequency? Does the ibis buffer have I-V and v-t curves or only I-V? My experience is that for most cases, the simulation and measurement do agree reasonably well. Generally it is the typical case but there have been times when either the slow or fast case agrees better.

Kai Keskinen

darmoni
Posts: 8
Online: User is Offline

11/15/2007 6:16 AM
Hi, thank you for the reply, yes, the BW of the scope is enough, i used an active probe and the setup of the simulation is done (post layout). i suspect the ibis but don't know how to run serie of tests to check it. thank you, Doron.

Kalevi2
Moderator
Posts: 69
Online: User is Offline

11/15/2007 6:32 AM
Go to SigXP and drive the buffer into the reference load at the maximum interface rate. Do the waveforms look OK? If the model fails this test, it is no good. This test does not necessarily mean the model is accurate though.

Kai Keskinen

darmoni
Posts: 8
Online: User is Offline

11/15/2007 6:37 AM
what do you mean by reference load? i did that i think by driving R // C to GND and put a dummy prob. the waves were OK at 200MHz

Kalevi2
Moderator
Posts: 69
Online: User is Offline

11/15/2007 6:49 AM
The data sheet should have timing reference fixture. It should be the same in the ibis file but is often not. Use a dummy probe with the timing reference fixture and see if it works at the max frequency.

Kai Keskinen

LanceWang
moderator
Posts: 1
Online: User is Offline

11/16/2007 7:36 AM
This is a widely asked question for the IBIS models for the SI works. Here are few things I like to mention: 1. As Kai mentioned, it is important using reference load to validate your models. More practically, you need to use a real application load (at least a close one) to do so including working voltage range. IBIS models often fall a part when it is used in corner cases. You need to make sure it is ok if your design is in those situations; 2. Be sure to set the stimulus correctly when you test in SigXP. I meant stimulus rising / falling edges and voltage levels. In SigXP GUI, it has been set in the certain ways to drive the buffer. Sometimes, you will have to set it up by manually to make sure it is what you want to use. 3. if you use more than one tools, includes Spectre, don't assume they always come out the same results. It might not be the case. You will have to validate it one by one. By the way, there is a website (www.ibiscp.org） - IBIS Certification Program. It has some sample validation reports. You may get some ideas from there. Lance Wang lwang@iometh.com

darmoni
Posts: 8
Online: User is Offline

11/18/2007 12:10 AM
Hi, thank you. please see my test that i do. thanks, Doron.
Attachment: Correlation_11_9_temp.xls

Posting to forums is available to community members only.
Login or Register